Case studies in Data Mining & Knowledge Discovery

Knowledge Discovery is a process

● Data Mining is just a step of a (potentially) complex sequence of tasks

KDD Process

Data Mining & KnowledgeDiscovery Process Models

● How to organize a complex analysis project?● Need of development (and documentation)

guidelines to – improve efficiency

– avoid delays

– maximize re-usability of the work

From KDD to CRISP-DM

● In the following we will see sample projects and introduce CRISP-DM 1.0

From KDD to CRISP-DM

6

AIR MILESa case-study on customer segmentation

From: G. Saarenvirta, “Mining customer data”, DB2 magazine on line, 1998http://www.db2mag.com/98fsaar.html

7

Application: customer segmentation

Given: Large data base of customer data

containing their properties and past buying records

Goal: Find groups of customers with similar

behavior

8

clustering in 3D

Data: points in the 3D spaceSimilarity: based on (Euclidean) distance

9

Customer clustering & segmentationtwo of the most important data mining methodologies used in marketinguse customer-purchase transaction data to track buying behavior create strategic business initiatives. divide customers into segments based on

"shareholder value" variables: customer profitability, measure of risk, measure of the lifetime value of a customer, retention probability.

10

Customer segmentsExample: high-profit, high-value, and low-risk customers

typically 10% to 20% of customers who create 50% to 80% of a company's profits

strategic initiative for the segment is retention

A low-profit, high-value, and low-risk customer segment may be also attractive

strategic initiative for the segment is to increase profitability

cross-selling (selling new products) up-selling (selling more of what customers currently

buy)

11

Behavioral vs. demographic segmentsWithin behavioral segments, a business may create demographic subsegments. Customer demographic data are not typically used together with behavioral data to create segments. Demographic (sub)segmenting is used to select appropriate tactics (advertising, marketing channels, and campaigns) to satisfy the strategic behavioral segment initiatives.

12

The Loyalty Group in Canada

runs an AIR MILES Reward Program (AMRP) for a coalition of more than 125 companies in all industry sectors - finance, credit card, retail, grocery, gas, telecom.

60% of Canadian households enrolled

AMRP is a frequent-shopper program: the consumer collects bonuses that can then

redeem for rewards (air travel, hotel accommodation, rental cars, theatre tickets, tickets for sporting events, …)

13

Data captureThe coalition partners capture consumer transactions and transmit them to The Loyalty Group, whichstores these transactions and uses the data for database marketing initiatives on behalf of the coalition partners. The Loyalty Group data warehouse currently contains more than 6.3 million household records 1 billion transaction records.

14

Before data miningThe Loyalty Group has employed standard analytical techniques Recency, Frequency, Monetary value

(RFM) analysis online analytic processing tools linear statistical methods

to analyze the success of the various marketing initiatives undertaken by the coalition and its partners.

15

Data mining project at AMRPGoal: create a customer segmentation using a data mining tool and compare the results to an existing segmentation developed using RFM analysis. data mining platform DB2 Universal Database Enterprise

parallelized over a five-node RS/6000 SP parallel system.

Intelligent Miner for Data (reason: has categorical clustering and product association algorithms which are not available in most other tools)

16

Data model~ 50k customers and associated transactions for a 12-month period

17

Data preparation“shareholder value” variables revenue customer tenure number of sponsor companies shopped at

over the customer tenure number of sponsor companies shopped at

over the last 12 months, recency (in months) of the last transaction

calculated by aggregating the transaction data and then adding them to each customer record

18

Data preparation (2)

Dataset obtained by joining the transaction data to the customer file to create the input for clustering algorithms84 variables = 14 categories of sponsor companies 3 variables per category 2 quarters (first two quarters of 1997)

19

Data cleansingmissing values

demographic data is usually categorical has a high % of missing values the missing values can be set to either

unknown or unanswered (if result of unanswered questions)

if a large portion of the field is missing, it may be discarded. In the case study, missing numeric values set to 0

20

Data transformation

Ratio variables. E.g.: profitability = profit / tenure

Time-derivative variables. E.g.: profit 2nd quarter - profit 1st quarter

Discretization using quantiles. E.g., break points at 10, 25, 50, 75, and 90.

Discretization using predefined ranges. E.g., those used in census

Log transforms. E.g., for very skewed distributions

21

Distributionsoriginal data

22

Distributionsdiscretized data

23

Before/after discretization

24

Clustering/segmentation methodology

25

IBM-IM demographic clustering

Designed for categorical variablesSimilarity index: increases with number of common

values on same attribute decreases with number of different

values on same attribute

# of clusters is not fixed a priori only upper bound set

26

IM Demographic clusteringbasic parameters: Maximum number of clusters. Maximum number of passes through the data. Accuracy: a stopping criterion for the

algorithm. If the change in the Condorcet criterion between data passes is smaller than the accuracy (as %), the algorithm will terminate.

The Condorcet criterion is a value in [0,1], where 1 indicates a perfect clustering -- all clusters are homogeneous and entirely different from all other clusters

27

… more parametersSimilarity threshold. defines the similarity threshold between two

values in distance units. If the similarity threshold is 0.5, then two

values are considered equal if their absolute difference is less than or equal to 0.5.

In the case study: maximum # of clusters: 9 maximum # of passes: 5 accuracy: 0.1

28

Input datasetdataset: all continuous variables discretized. input variables :

# of products purchased over customer’s lifetime # of products purchased in the last 12 months Customer's revenue contribution over lifetime Customer tenure in months Ratio of revenue to tenure Ratio of number of products to tenure Region Recency Tenure (# of months since customer first enrolled in

the program).

29

Input dataset

Other discrete and categorical variables and some interesting continuous variables were input as supplementary variables:

variables used to profile the clusters but not to define them.

easier interpretation of clusters using data other than the input variables.

30

Output of demographic clustering

31

Visualization of clustershorizontal strip = a clusterclusters are ordered from top to bottom in order of sizevariables are ordered from left to right in order of importance to the cluster, based on a chi-square test between variable and cluster ID. other metrics include entropy, Condorcet criterion, and database order.

32

Visualization of clustersvariables used to define clusters are without brackets, while the supplementary variables appear within brackets. numeric (integer), discrete numeric (small integer), binary, and continuous variables have their frequency distribution shown as a bar graph.

red bars = distribution of the variable within the current cluster. gray solid bars = distribution of the variable in the whole universe.

33

Visualization of clustersCategorical variables are shown as pie charts. inner pie = distribution of the categories for the current clusterouter ring = distribution of the variable for the entire universe.

The more different the cluster distribution is from the average, the more interesting or distinct the cluster.

34

Qualitative characterization of clusters

Gold98 is a binary variable– indicates the best customers in the

database, created previously by the business using RFM analysis.

The clustering model agrees very well with this existing definition– most of the clusters seem to have

almost all Gold or no Gold customersConfirmed the current Gold segment!

35

Qualitative characterization of clustersOur clustering results not only validate the existing concept of

Gold customers, they extend the idea of the Gold customers

by creating clusters within the Gold98 customer category.

A platinum customer group

Cluster 5 almost all Gold98 customers, whose

revenue, bonus collected lifetime to date, revenue per month, and lifetime to date per month are all in the 50th to 75th percentile.

36

Qualitative characterization of clusters

Cluster 3: no Gold98 customers. Its customer revenue,

bonus collected, revenue per month, are all in the 25th to 50th percentile.

Cluster 5: 9 %of the population. revenue, bonus collected are all in the 75th

percentile and above, skewed to almost all greater than the 90th percentile.

looks like a very profitable cluster

37

Detailed view of cluster 5

38

Profiling clustersGoal: assess the potential business value of each cluster quantitatively by profiling the aggregate values of the shareholder value variables by cluster.

39

Profiling clustersleverage = ratio of revenue to customer.

cluster 5 is the most profitable cluster. as profitability increases, so does the

average number of products purchased. product index = ratio of the average

number of products purchased by the customers in the cluster divided by the average number of products purchased overall.

customer profitability increases as tenure increases.

40

Business opportunitiesBest customers in clusters 2, 5, and 7. :

indication: retention

clusters 2, 6, and 0 indication: cross-selling by contrasting with

clusters 5 and 7. Clusters 2, 6, and 0 have a product index close

to those of clusters 5 and 7, which have the highest number of products purchased.

Try to convert customers from clusters 2, 6, and 0 to clusters 5 and 7. By comparing which products are bought we can find products that are candidates for cross-selling.

41

Business opportunities

Clusters 3 and 4 indication: cross-selling to clusters 2,

6, and 0

Cluster 1 indication: wait and see. It appears to

be a group of new customers

Cluster 8 indication: no waste of marketing

dollars

42

Follow-upReactions from The Loyalty Group

visualization of results allowed for meaningful and actionable analysis.

original segmentation methodology validated, but that refinements to the original segmentation could prove valuable.

decision to undertake further data mining projects, including

predictive models for direct mail targeting, further work on segmentation using more detailed

behavioral data, opportunity identification using association algorithms

within the segments discovered.